How to create recommendation systems that are suitable for deployment in production environment for an ecommerce giant?

I am making a recommendation model for an ecommerce client that has huge number of products of various categories. Product data set can be considered similar to that of Amazon. For now, I am starting with simpler models like content based and collaborative filtering and targeting cross-sell and upsell use cases like Similar Products to current item, Users who viewed this item also viewed, Users like you also liked etc.

I am wondering if I need to create separate models for all different categories (like apparels, electronics, food, sports etc.) and build separate APIs for them. I've seen POCs and examples of recommendation systems(mostly on MovieLens Dataset) but those lack readiness for production environment.

Since there are many different categories of products in ecommerce, generating a separate recommendation model for each category and then maintaining it would be cumbersome. For e.g., if there are 1000 different categories/subcategories of products, I don't want to create 1000 recommendation models for each of them. Is there any optimized approach for creating/managing these models?